CBL - Campus del Baix Llobregat

Projecte llegit

Títol: Study and application of machine learning techniques to the deployment of services on 5G optical networks

Estudiants que han llegit aquest projecte:

Director/a: SPADARO, SALVATORE

Departament: TSC

Títol: Study and application of machine learning techniques to the deployment of services on 5G optical networks

Data inici oferta: 09-07-2018     Data finalització oferta: 09-03-2019


Estudis d'assignació del projecte:
    MU MASTEAM 2015
Tipus: Individual
 
Lloc de realització: EETAC
 
Paraules clau:
5G, Machine Learning, Optical Networks
 
Descripció del contingut i pla d'activitats:
The aim of this TFM is to design and deploy the environment for the application of Cognition-based techniques (e. g., Big Data and Machine Learning) to the control and management of optical networks supporting 5G-based services. Such environment will have to collect monitoring information from the transport network and do a suitable processing to infer the necessary performances, in order to fulfil the SLAs associated to the service offered. To do this, the candidate will have to investigate a series of tools destined to the data collection (e. g., SkyDive, Kafka, etc.) and to the application of machine learning techniques (e.g., Spark, Hadoop, ...), as well as with paradigms of management and control of SDN/NFV optical networks.
 
Overview (resum en anglès):
The vision of the future 5G corresponds to a highly heterogeneous network at different levels; the increment in the number of services requests for the 5G networks imposes several technical challenges.
In the 5G context, in the recent years, several machine learning-based approaches have been demonstrated as useful tools for making easier the networks’ management, by considering that different unexpected events could make that the services cannot be satisfied at the moment they are requested. Such approaches are usually referred as cognitive network management.
There are too many parameters inside the 5G network affecting each layer of the network; the virtualization and abstraction of the services is a crucial part for a satisfactory service deployment, being the monitoring and control of the different planes the two keys inside the cognitive network management.
In this project it has been addressed the implementation of a simulated data collector as well as the study of several machine learning-based approaches. This way, possible future performance can be predicted, giving to the system the ability to change the initial parameters and to adapt the network to future demands.

© CBLTIC Campus del Baix Llobregat - UPC